ELECTRONIC BULLETIN BOARDS AND "PUBLIC GOODS" EXPLANATIONS OF COLLABORATIVE MASS MEDIA Sheizaf Rafaeli School of Business Administration Hebrew University of Jerusalem Mount Scopus, 91905, Jerusalem ISRAEL Tel. 02-883106 Online: KBUSR@HUJIVM1.bitnet and Robert J. LaRose Department of Telecommunication Michigan State University East Lansing, MI 48824-1212 USA Tel. 517-355-4528 March, 1992 Dr. Rafaeli is an Assistant Professor at the School of Business Administration, the Hebrew University of Jerusalem. Dr. LaRose is an Associate Professor at the Department of Telecommunication, Michigan State University. The authors wish to acknowledge their gratitude for support provided by Michigan State University's Communication Technology Laboratory, the Northern Telecom University Interaction Program, and the Rekanati Fund. ELECTRONIC BULLETIN BOARDS AND "PUBLIC GOODS" EXPLANATIONS OF COLLABORATIVE MASS MEDIA ABSTRACT Collaborative mass media are a new type of mass communications medium in which the audience acts both as the source and the receiver of the message. Theories of discretionary data base contributions and critical mass theory offer parallel, but in some ways distinct, explanations for the success of collaborative media. The present research compared the predictions of these two perspectives in the context of a national survey of public electronic bulletin board systems. The study documented the nature and extent of electronic bulletin board use and compared seemingly conflicting predictions about the success of collaborative media based on the two theoretical perspectives. File contribution levels and system adoption rates were both found to be directly related to a measure of symmetry in user participation. Content diversity was directly related to contribution levels, but not to overall adoption levels. The results provided limited support for discretionary data base theory. ELECTRONIC BULLETIN BOARDS AND "PUBLIC GOODS" EXPLANATIONS OF COLLABORATIVE MASS MEDIA Collaborative mass media systems, in which the audience is also the primary source of media content as well as its receiver, represent a new and significant departure from conventional mass media forms. They expand the very definition of mass media, from "one to many" to "many to many" communication. In the past, audience-generated mass media content has invariably been subjected to a considerable degree of editorial control and has generally constituted a relatively small percentage of total message system content. Contributions came from a such a small number of audience members that the participants were more properly regarded as symbols of the community of interest rather than a true embodiment of it. "Letters to the Editor" columns in newspapers and magazines and radio and television call-in programs are familiar examples. In contrast, new collaborative media mass forms rely almost exclusively upon contributions from a wide cross section of audience members, often with minimal editorial control. Electronic bulletin board systems, audiotext "chat lines" and videotext fora are three examples of emerging interactive mass communication systems in which the audience is also the source of the content. To understand these new media forms, we must understand the factors which prompt audience members to make the contributions needed to make them viable. Participation in such media is paradoxical, after all. They are available to all members in a community of interest, regardless of whether the individual consumers contribute themselves, and consumption of the medium does not diminish what is available to anyone else. Since no single individual is required to contribute to the content of the message system in order to benefit from it, the "rational consumer" might be expected to adopt the strategy of consuming the medium without making contributions. Yet, if no one contributed there would soon be no medium to consume! It thus seems that interactivity itself elicits behavior that transcends the predictions of models of "rational" economic consumption of information (Rafaeli, 1990; Rafaeli and Ritchie, 1991). More formally, this is the dilemma of so-called "public goods" (Barry & Hardin, 1982). How much, why, and when do people contribute to interactive mass media? How can we predict, and eventually increase, sharing of ideas, information, intellectual property, and cognitive work among peers in communities defined by shared communication environments? Two theoretical perspectives have emerged recently which promise to extended our understanding of interactive mass media forms which rely on the participation of their audiences. Thorn and Connolly (1987, 1990) developed a theory of discretionary data bases from the notion of public goods and used it to predict rates of contribution to stores of information shared within an organization. In a series of laboratory experiments, they found that the lower the perceived value of the information and the less symmetrical the benefits to users in a group, the lower the contribution levels. Furthermore, individuals contributed less as participation costs and group size increased. However, these results were limited to extremely small (four or eight person) groups and took place in a laboratory setting that more closely replicated small workgroup communication than a mass media context. Critical Mass theory (Markus, 1987, 1990) takes a parallel approach, also predicated on the public goods dilemma, which attempts to explain the growing adoption of interactive media in a community of interest until a state of near total participation, or universal access, comes to exist. The theory predicts that the chances of attaining universal access are inversely related to the resource contributions -- in terms of skill, effort or cost -- required of users. On the other hand, the greater the heterogeneity of interests and resources found among members of a community, the greater the chance of achieving universal access. Task interdependence, centralization of resources, group size and geographic dispersion are hypothesized to be directly related to heterogeneity and hence to adoption. Critical mass was initially conceptualized with the problem of adoption of interpersonal communication media -- such as the telephone and electronic mail -- in mind, but the public goods argument on which it is based would seem to apply equally well to collaborative mass media systems. The two theories thus share common underlying assumptions about human behavior in contexts in which collaborative behavior is required to create a communications medium. They both address the same general problem of explaining participation in interactive media that are subject to the dilemma of public goods. Both paradigms have matching predictions for one key set of variables relating to the demands placed on the user. The more effort, skill, or monetary cost involve, the less the participation levels. The two approaches also differ in some important aspects. Their independent and dependent variables, units of analysis and the nature of their predictions are somewhat distinct. Thorn and Connolly's theory of discretionary data bases focuses on individual contribution levels as the dependent variable of interest, operationalized in terms of the percentage of transactions in which subjects choose to contribute information. For Markus, the dependent variable is the collaborative adoption rate in a community of interest, or the percentage of the community that has adopted the innovation in question. At a purely operational level, the distinction is simply that critical mass focuses on the percentage of a community that uses a collaborative medium, while the discretionary data base paradigm examines the subset of users who also make contributions to the content of the medium. Presumably, there is also a difference in the time order of the two outcomes. Initial adoption behavior -- including the purchase of necessary hardware and software, obtaining the necessary access codes, logging onto the system for the first time, etc. -- can be expected to precede contribution behavior. This distinction may in turn explain the differences in independent variables. The value of information and the symmetry of benefits are qualities that can perhaps best be sampled after initial access has been obtained through the process of adoption, hence their absence in critical mass theory. Still, it is reasonable to expect that users form perceptions of these qualities even before initial adoption, so that they could be expected to influence the adoption decision as well. Note that the critical mass approach makes the opposite prediction regarding the role of group size. If larger communities are also more heterogeneous, then critical mass theory predicts that universal access is more likely to be achieved in larger communities. Indeed, in this view participation is not just directly related to group size, it is exponentially related to the number of participants (cf. Markus, 1987). Thorn and Connolly (1990, p. 227) argue that large systems reduce the expectation of reciprocity, and consequently decrease contribution levels. Common sense offers a more homely construction: the more users there are, the more each user can safely assume that "someone else will do the job." This could even be a positive factor when making an adoption decision. When considering whether to adopt a new collaborative medium potential users may find larger systems more attractive precisely because they seem to require a lesser investment of resources -- in terms of their own contributions -- than smaller systems do. For the purposes of the present research, one possible way to resolve the distinctions between these two public goods perspectives is to consider them both to be about the factors that make collaborative media successful. The two perspectives merely focus on different indicators of success. One defines success in terms of the proportion of a community of interest that has adopted the innovation, while the other focuses on the contribution levels of the participants. Other outcome variables come to mind, as well. Since collaborative media also have mass media characteristics, it is natural to consider conventional mass media outcome measures, such as the amount of usage of the system. In the case of highly diverse and volatile media, such as the electronic bulletin board systems considered here, system longevity is another appropriate outcome measure. Electronic Bulletin Boards Electronic bulletin boards marry the twin interests of mass and interpersonal communication (Rafaeli, 1986). The first electronic bulletin board is believed to have gone online in Chicago, in early 1978 (Anis, 1991). Today, bulletin board systems are operated from mainframes as well as stand-alone personal computer platforms. Some have a narrow, well-defined areas of interest, while others serve as a general meeting place for anyone with a personal computer and a modem. Most boards are based on audience participation (Rafaeli, 1986; Steinfield & Fulk, 1988), which makes them an interesting example of collaborative mass media. The popular press has been providing wide, if jaded coverage (e.g. Alexander, 1991; Bromberg, 1991; USA Today, 1990). The medium has also spawned its own trade publications, Boardwatch and Plumb. Between ten and thirty thousand different systems with dozens to hundreds of participants each are currently in operation in North America. Many of these are operated by and for the benefit of closed groups. The number of publicly open boards is not known, however such boards clearly number in the thousands. The publicly open boards bring the public goods issues into particularly sharp focus. Closed boards operating in an organizational context can easily resort to various forms of coercion or to monetary incentives to maintain contributions. Indeed, Thorn and Connolly (1990, p. 226) recommend the latter strategy to corporate communication managers. However, the introduction of these external variables represents a serious departure from the basic public goods dilemma, which assumes free choice in the provision of information. In contrast, public boards are more nearly a "pure" case of collaborative mass media in which the exchange of information is in itself the reason for being. Most of the attention paid to bulletin board systems by communication researchers has so far focused on the characteristics of demand and consumption -- "taking", "getting", or "Uses and Gratifications" (Chesebro, 1985; Danowski, 1982; Foulger, 1990; Garramone et al. 1985; 1986a; 1986b; Hellerstein, 1989; Rafaeli, 1986; Rogers and Rafaeli, 1985; Rogers, 1990; Swift, 1989). However, most published work is based on case studies of single systems at single points in time, offering little opportunity to assess variables which affect participation across systems. The present research furthers the study of bulletin boards along the axes of the "hypercube" described in Figure 1. As noted, most research to date has been concentrated in one far corner. They are one-shot case studies, focusing on private or institutional contexts, devoted to special purposes, and often experimental in nature. [Figure 1 about here] Furthermore, bulletin boards represent interactive media which require contribution as well as consumption of their users. Studies which focus on usage tend to examine only the "receiving end" of the interaction. The public goods concept directs our attention to the reciprocal "giving" behavior involved in the operation of the medium. In the present study, the focus was on the reciprocity of communication. The units of analysis were communities of participants, rather than individual users. A Model of Collaborative Mass Media Behavior The present research utilized a model of collaborative mass media behavior applied to electronic bulletin board systems. It draws upon the conceptualizations of discretionary data bases and critical mass described above. The dependent variables of interest all relate to the success of electronic bulletin board systems, described in terms of adoption (after Markus, 1990) contribution levels (after Thorn and Connolly, 1990), usage and longevity. The independent variables reflect the concerns about resource requirements, heterogeneity and reciprocity found in public goods explanations of collaborative media. In order to assure the external validity of the findings, an effort was made to reconstrue management policies identified by "real world" bulletin board managers into variables which are theoretically meaningful for collaborative mass media subject to public goods explanations. These included "exchange rates" instituted to police the traffic of files, the diversity of content on the board, and the degree of communication reciprocity maintained. H1: Bulletin board success will be negatively related to access restrictions placed on users. Many bulletin board system managers apply downloading restrictions on users in an effort to encourage contributions. One common type of restriction is an upload to download ratio; e.g. one file must be uploaded for every ten that are downloaded in order to maintain access privileges. Another common restriction is the imposition of time limits on calls in an effort to keep incoming lines available for other users. Such practices increase contribution costs in Thorn and Connolly's terms, or represent a substantial "communication discipline requirement" in critical mass terms, and should inhibit both adoption and contribution levels. Of course, the imposition of user fees is perhaps the most obvious example of a contribution cost or communication requirement that can be expected to limit bulletin board success. H2: Bulletin board success will be positively related to the diversity of content available. Thorn and Connolly (1990, p. 224) specify the quality of information, in terms of its value to the user, as an important determinant of contribution behavior. At the system level, we hypothesize that the more diverse the content is, the greater the likelihood that users will find something of value and will in turn be prompted to make contributions in order to elicit further quality contributions. Critical mass theory does not treat message system content as a variable, an important oversight, perhaps. However, we can deduce from Markus' proposition about heterogeneity of interests that the more diverse the content, the higher the adoption levels. The issues of diversity has a somewhat unique meaning in the world of electronic bulletin boards. Bulletin board systems originated as communications media for computer hobbyists and many maintain this focus, therefore a meaningful indicator of diversity is the proportion of content devoted to non-computer related topics. H3: Group size will be negatively related to contribution levels but positively related to other measures of bulletin board success. This hypothesis applies Thorn and Connolly's prediction about contributions and group size to a more externally valid context. Following critical mass theory, the opposite prediction is made for adoption levels and other outcome measures of bulletin board success. H4: Symmetry in contribution levels will be positively related to bulletin board success. Thorn and Connolly state that asymmetries in contribution costs will reduce contributions. This means that boards in which a large proportion of users make contributions to the collaborative medium should be more successful than those in which only a few users contribute. Thorn and Connolly also predict that either too many contributors or too many non-contributors will reduce contribution levels. There needs to be a balance between the two in their view. However, unlike the closed organizational systems that Thorn and Connolly have studied, collaborative mass media are open systems in which the "problem" of too many contributors could be offset through the addition of new users to the system. In view of this, we hypothesized that collaborative mass media would demonstrate a direct relationship with symmetry in contributions. Markus (1987, p. 500) also notes that usage falls off in the absence of reciprocity. The proposition about reciprocity in critical mass theory does not require that the exchange be equal, only that some number of others reciprocate, however. RESEARCH METHOD Sampling A list of 7600 bulletin boards was constructed from three sources available in the public domain. These lists were originally put together by aficionados, producers of bulletin board software packages, and operators of bulletin boards. The collated list was checked for duplication, resulting in a list of 4800 telephone numbers of electronic bulletin boards. This list thus constituted the universe of operating, publicly accessible boards whose operators had attempted to list in national directories. A random sample of 500 boards was obtained from the population list, using a random start and fixed skip interval. Procedure A computer readable questionnaire, consisting of 37 open and closed ended items, was created. The questionnaire was uploaded to the system operators in the sample by calling their user- access lines. In all, 293 questionnaires were uploaded. Up to five callbacks were made to each board in an attempt to upload the questionnaire. Most calls were made in the late night and early morning hours, over a period of two months. The remaining 207 numbers included disconnected numbers, constantly busy numbers, no answers, wrong numbers, voice lines and fax machines. In addition, about 5 percent of the numbers in the original list are not included in the final sample because the systems were closed to the public. The questionnaire requested system operators to enter their answers in the spaces provided in the computer-readable questionnaire form and to post their completed questionnaire on their bulletin board. Alternatively, operators mailed or faxed their responses. Approximately three-quarters of the responses were collected by calling back the systems. Two call back attempts were made for each uploaded questionnaire. There were 126 valid responses, for a completion rate of 42 percent of the public boards contacted. Twelve percent of the boards refused the survey. Operational Measures Five dependent measures relating to the success of electronic bulletin board systems were defined: Contribution Level was operationally defined as the ratio of files contributed, or "uploaded," in a week to the total number of weekly file transactions, including files uploaded and files taken, or "downloaded" by users. Respondents were asked, "How many files UPLOADED[DOWNLOADED] in an average week?" and these estimates were used to compute the ratio. This measure corresponds closely to that in the Thorn and Connolly (1990) experiments, in which the percentage of experimental transactions in which information was contributed was the dependent variable. Adoption Rate was the ratio of regular callers ("How many REGULAR callers who call at least once a week") to the total number of users ("How many DIFFERENT callers do you estimate you have.") It measures the adoption of the system among the community of interest defined by each board. Longevity was the answer to the question, "How many months ago did this board first go on line?" Usage was the "Average number of calls on an average day." Independent variables relating to user restrictions, diversity, group size and communication symmetry were also defined: Ratio Restriction was a dichotomous variable, scored 1 if the board operator had upload-download restrictions, 0 if not. Replies to the question "What are your policies about upload/download ratios and how have they changed over time" were coded in this fashion. Time Restriction was the response to the question, "What is the online time limit?" Thus, in this case, the higher the number, the less restrictive the policy. Fee Requirement was scored 1 if the operator indicated that any revenues were derived from recurring user fees or usage fees, 0 if no user fees were required. Diversity was defined as the response to the question, "What percent of the content on your board would you say is NOT computer related?" Group Size was the total size of the community of interest ("How many DIFFERENT callers do you estimate you have?") Symmetry was the percentage of board users characterized by the board operator as being either "givers" ("who share knowledge and resources") or "exchangers" ("participating in a fair exchange"), as opposed to "Takers" ("who are looking for 'something for nothing'") Additional questions about the characteristics of boards, their activity levels and management practices were also included in the survey. Open ended-questions about purpose ("How do you define the purpose of your board?"), scope ("How do you describe your target community?") and success factors ("What do YOU think makes a bulletin board system successful?") were included to further discern the nature of bulletin board activity. RESULTS Usable responses were received from 126 boards located in forty- five different states. Some of the boards were quite long-lived, including one that was eleven years old at the time of the survey, and twenty-five percent of those surveyed were over five years old. Thirteen percent had been in existence for only a year or less. The average longevity was about three and a half years (41.6 months). On average, thirty-eight percent of the weekly file transactions involved the contribution of files from users. An average of eighteen percent of the boards' user universe (defined as the number of different callers) are active users, placing calls on a weekly basis. Slightly over fifty calls per week are made to the typical board in the survey, but the range is quite wide. One of the boards reported 2000 weekly calls. The average user community served included nearly 900 (886.9) different users. The boards surveyed were very diverse in terms of the nature and scope of their target audiences. The boards included in the sample ranged from general purpose systems available to all computer users (39%) to highly specialized systems dedicated to specific groups. Twenty percent of the system operators said that their boards were either national or international in scope. The specific audiences cited included computer professionals, handicappers, children, lawyers, ham radio operators, AIDS professionals and victims, physics researchers. Sixteen percent of the boards identified upper class (i.e., high income or highly educated) individuals as their target audience. Bulletin board systems are defined as much by the nature of their content as by the characteristics of their audiences. Some of the more highly specialized systems in the sample included public boards operated by large corporations and service organizations and product support bulletin boards for software companies. Of course, all board users share an interest in computing, since a personal computer or terminal is required for access. On average, about two-thirds of board content is devoted to computer-related matters, but only nine percent of the boards were exclusively computer related. In contrast, twenty-two percent of the boards said that half or more of board content was unrelated to computer topics. Abuse is an important issue for board operators. Common abuses include evasion of rules against excessive downloading and lengthy sessions that tie up incoming lines, the uploading of "junk" files and the placement of obscene or abusive messages on the boards. Over two-thirds of the boards in the sample reported instances of abusers. The most common policy for dealing with such abuse was the termination of access rights. Fewer operators reduced access rights, and only about half limited downloading. The least popular strategy was notification of the police, reported by one in seven operators. A wide variety of board management policies are in effect to forestall such problems. Twenty-three percent of the boards said they charged either recurring or per transaction fees to users. Forty-one percent enforce some form of download ratio in an attempt to curb "selfish" users who merely take files without offering up files of their own. Time limits were in effect for all but six percent of the boards. The average maximum length of call allowed was about an hour (55.6 minutes). The longest time limit was two hours. In fact, 44 percent of all boards surveyed had a time limit set at exactly 60 minutes. Furthermore, most board operators award access time by user status. Less than half credit additional time based on upload history, and about a third allow users to purchase access time. Other access policies included preferential treatment for visiting operators from other systems, time of day differentials, and the award of more time upon request. Another mark of successful board management is the ability to develop active participation. When asked to classify their users as "Givers", "Exchangers", or "Takers", system operators classified more of their users as "Takers", than either of the other two categories. The median percentage assigned to "Takers" was 50 percent, while the median for each of the other, 'less selfish' categories was 20 percent. The "art" of running a bulletin board is still being defined. When asked to describe factors in the success of bulletin board systems, the most frequently cited factor was the responsiveness of the system operator to user needs, which was mentioned by twenty-three percent of the respondents. The quantity of files available and the quantity of user contributions were each cited by nineteen percent of the respondents. Table 1 displays means, standard deviations and correlation coefficients for the dependent and independent variables used to test the hypotheses for the present study. The dependent measures of system success were generally not very highly correlated. The only significant correlation among the measures of system success was a negative correlation between contribution levels and system longevity (Pearson r=-.21, p < .05). Accordingly, the patterns of relationships with the independent variables were considered separately for each of the indicators of system success. Contribution levels were not significantly related to the presence of upload-download ratio restrictions, time restrictions, or the imposition of user fees. The relationships to transaction ratios, time restrictions and fee policies were in the directions hypothesized, even though they were not significant. Content diversity was directly related to contribution levels, as hypothesized (r=.27, p < .01), as was symmetry in contribution levels (r=.22, p < .05). Group size was not significantly related to contribution levels, although again the relationship was at least in the direction expected. Adoption levels, defined in terms of the proportion of regular users among the total community of users, were negatively related to group size (r=-.33, p < .01), the opposite direction of what was predicted from critical mass theory. Symmetry in contribution activity was positively related to adoption levels (r=.30, p < .01), as expected. The other hypotheses were not supported and the directionalities of the relationships observed were generally opposite those predicted. The exception was that leniency in access time restrictions was positively related to adoption levels, which was the direction hypothesized. The longevity of the boards surveyed was positively related to group size, as predicted (r=.40, p < .01), but the other hypothesized relationships were not supported. Only the prediction about content diversity was in the anticipated direction. Finally, weekly usage levels were directly related to group size as hypothesized (r=.27, p < .01), but were positively correlated to the existence of a fee policy (r=.19, p < .01), contrary to expectation. The other relationships were not only not significant, but tended to be in the opposite of their predicted directions. DISCUSSION The exchange of files and messages occurring on electronic bulletin board systems has a mass communication flavor to it. It has been occurring on a fairly large scale for well over a decade. Unlike traditional mass media vehicles which involve one-to-many dissemination of content, bulletin boards are collaborative efforts defined by a many-to-many relationship. The members of the community of interest participate in the creation of the message. Such participation is of both practical and theoretical import. It appears that structural characteristics of collaborative mass media systems seem to be more critical to their success than specific management policies applied by system operators. The diversity of content and the symmetry of exchange between participants were the most important of the factors derived from public goods theories which predict bulletin board system success. The day-to-day operating restrictions placed on users -- including upload ratios, access time restrictions and user fees -- generally had little relationship to measures of board success. The above statement must be heavily qualified according to which measure of success is in question. The predictions of public goods theory best explained contribution levels, but did not seem to be a "good fit" for the other outcome measures of adoption, longevity and usage. The results were not very robust even for contribution levels. Perhaps there are distinct processes which uniquely explain these different aspects of collaborative mass media systems, apart from public goods explanations. It seems that the predictions of public goods theories need further refinement when applied to collaborative mass media. In somewhat of a challenge to Public Goods theory, and mild support of Critical Mass theory, symmetry in user contribution behavior was related positively to both contribution and adoption rates. It may be that the open-ended nature of public collaborative media makes them a different case from collaborative media that exist in closed group settings, such as in organizations. When a public board is the beneficiary of exceptional contributions from users, there is a ready supply of new users who will be attracted and offset any problems with the "over-supply" of information. The present results perhaps indicate a need to re-assess the importance of symmetry in contributions from user groups in open media systems. The picture emerging from these data is of a more complex world than predicted by either theory. There were small-scale, rationally managed, tightly economical (or exchange-governed) boards, where the public goods of information were traded. At the same time, many other boards exist and survive, where the emphasis is on diversity, numerous callers, and generalized reciprocity expectations rather than tightly managed quantitative exchange rates. In future research it would be worthwhile to re-conceptualize and test some other system management issues in public goods terms. In the present sample of system operators, their own responsiveness, the quantity of files available and the quantity of user contributions were deemed the most important factors in success. The operator, as "the master user," has the ability to control the symmetry of the exchange through her own contributions to the board. This might be especially true of smaller boards, in which an active system operator could affect the gross exchange rate in meaningful ways. In such cases, the expertise of the operator and the centralization of resources represented by that person could create the "heterogeneity of resources" which critical mass theory holds is an important driver of universal access. The quantity of files available also has meaning in public goods theories. The larger the file library, the more likely that "heterogeneity of interests" may be obtained, in critical mass terms, driving user adoption. Thorn and Connolly's theory of discretionary data base contributions would seem to make the opposite prediction with respect to contribution levels: Faced with a wide array of valuable files, individual users might be discouraged from making contributions of their own. The chances of inducing others to contribute by offering files that only make a marginal addition to an already expansive data base would seem to be slight, so contribution behavior should be inhibited by the sheer volume of files available. While on the topic of system content, this seems to be another area worthy of further consideration by public goods theorists. In the current research, the diversity of content was related to contribution levels, although not to adoption rates. There is room for much refinement of the content diversity variable, such as defining it in relation to the number of different topics carried and the number of files devoted to each. The quality of the content should also be an important factor, although it is difficult to imagine a satisfactory measure of quality that could be assessed at the system level. The final "success factor" identified by the system operators -- the quantity of files contributed -- suggests that there may be an order of precedence among the various success factors examined in the present research. A healthy contribution rate may be a precondition of universal access within a community of interest. The "Community of Interest" Problem Overall, critical mass theory found only slight support, correctly predicting that adoption levels would be related to reciprocity between users. An important caveat on the present findings as they apply to critical mass theory is that the dependent measure of adoption used here -- the ratio of regular users to the total number of different callers -- was not entirely satisfactory. For collaborative mass media systems, it is difficult to specify "community of interest," the denominator used to calculate the adoption rate in critical mass theory. Part of the problem is that electronic bulletin board use entails a cluster of innovations, not a single innovation. One must first adopt home computer technology, then data communications and then the behavior of searching bulletin board systems before one is in the community of interest. Further restrictions could be placed on the definition of the community according to the thematic content of the boards and the characteristics of their target audiences. The community of interest could perhaps then be defined as "People with personal computers and computer modems who access bulletin board systems and who are interested in the topics that the board specialized in." In any event, the system operators who participated in the study apparently had no way to meaningfully estimate the size of their communities of interest. When asked to identify their target populations, most seemed to base their estimates on the total population of their local calling areas. The use of these estimates did not seem appropriate as they would have made the adoption rates too dependent upon a third variable: the size of the board's municipality of origin. At least one of the contrary findings of the present study may have been due to the nature of the adoption rate measure that was used. The adoption rate was a simple transformation of the group size variable. Group size was in the denominator of the adoption rate measure, so it was no surprise that the two were negatively correlated. The definition of the community of interest is likely to remain a thorny issue in future investigations of collaborative mass media that are publicly offered. By their very nature, collaborative mass media have self-defining communities of interest. The very act of accessing a bulletin board system is what defines one both as a member of the community of interest and as an adopter of the service. For Further Research This study had some important limitations that should be addressed by future research. We restricted ourselves to public boards, and the sampling strategy included only boards that were in operation at the time of the study. A resulting limitation of the sampling scheme was that only successful boards participated, restricting the range of validity of our claims. Perhaps we failed to note the effect of critical mass for the simple reason that systems who failed to achieve critical mass had already failed and "selected themselves" out of the sample. The results reported here may merely indicate that the bulletin board system medium has not yet reached the "stable state" cited as an important milestone in critical mass theory. Future studies should employ time series methods to examine over-time effects of the independent variables on community adoption rates within individual bulletin board systems. This study was also subject to many of the problems of self- report surveys. Some of the responses, most notably estimates of financial resources and growth patterns, were so riddled with apparent inconsistencies that we chose not to include them. However, some of the data was at least machine-generated, if still self-reported. In many cases the software used to operate the boards also provided objective statistics about specific items including the number of callers and file activity levels. In several instances, system operators even appended computer files of usage statistics that were automatically generated by their operating software. An innovative approach to overcoming these limitations would be to create "experimental boards," in which variables of board management policy could be carefully controlled and manipulated and their effects on system success evaluated with machine- generated data. Such an approach could achieve a satisfactory degree of experimental control in a context that has a high degree of external validity, while eliminating the possible biases of self- reported data. There would also be the opportunity to survey board users so that key assumptions about the linkage between individual and system-level variables (e.g., that users are capable of perceiving symmetry in contribution levels) could be verified. In conclusion, the documentation of the type of collaborative media behavior represented by bulletin board systems remains an important topic. Bulletin boards have outlived many of the failed videotext experiments, so far without attracting as much attention. The focus on the unit of analysis of entire communities of users is also important methodologically. The study of interactive technologies needs to proceed beyond the case-study level in hopes of discerning the factors that lead to the success or failure of collaborative mass media systems. TABLE 1 DESCRIPTIVE STATISTICS AND PEARSON PRODUCT-MOMENT CORRELATION MATRIX _________________________________________________________________ _________________________________________________________________ _ VARIABLE Mean S.D. 2 3 4 5 6 7 8 9 10 _________________________________________________________________ _________________________________________________________________ _ 1 Contributions .38 .29 .06 -.21* -.11 -.13 .00 -.03 .27+ -.07 .22* 2 Adoption .18 .15 -.12 -.07 .11 .14 .00 -.04 -.33+ .30+ 3 Longevity 41.61 26.17 .11 .12 -.15 .07 .11 .40+ -.09 4 Usage 51.69 183.00 .12 .00 .19* - .02 .27+ -.07 5 Ratio Restricted .41 .49 -.01 .38+ -.18 -.09 .09 6 Time Restricted 55.52 25.59 - .06 -.02 -.23* .15 7 Fee Restricted .23 .42 -.26+ .06 -.23* 8 Diversity 31.58 30.49 .14 .15 9 Group Size 886.91 1198.85 -.22* 10 Symmetry 51.39 31.33 _________________________________________________________________ ________________________________________________________ Note: Table entries are Pearson Product-Moment correlations. * p < .05 + p < .01 REFERENCES Alexander, M. (1991) Suit seeks to define user rights, ComputerWorld, XXV:18, May 6, 1. Anis, N. (1991) Interview with Ward Christensen the first BBS operator, BoardWatch Magazine, April, p. 14-22. Barry, B. & Hardin, R. (Eds.) (1982). Rational man and irrational society? Beverly Hills, CA: Sage. Bromberg, C. (1991) In Defense of Hackers, New York Times Magazine, April 21st, p. 44. Chesebro, J.W. (1985) Computer mediated interpersonal communication. In B.D. Ruben (Ed.) Information and Behavior, Vol. 1, 202-222. New Brunswick, NJ: Transaction Books. Danoswski, J.A. (1982) Computer mediated communication: A network-based analysis using a CBBS conference. In M. Burgoon (Ed.) Communication Yearbook 6. Beverly Hills: Sage, pp 905- 924. Foulger, D. (1990) Medium as Process: The structure, use, and practice of computer conferencing on IBM's IBMPC computer conferencing facility. Ph.D. Dissertation, Temple University. Garramone, G.M., Harris, A.C., and Anderson, R. (1986a) Uses of political computer bulletin boards. Journal of Broadcasting & Electronic Media, 30(3), 325-339. Garramone, G.M., Harris, A.C., & Pizante, G. (1986b) Predictors of Motivation to use computer-mediated political communication systems. Journal of Broadcasting & Electronic Media, 30(4), 445- 457. Hellerstein, L. (1989) Creating social reality with computer- mediated communication. Ph.D. Dissertation, Univ. of Massachusetts. Markus, M.L. (1987) Toward a "Critical Mass" theory of interactive media: universal access, interdependence and diffusion. Communication Research, 14(5), pp 491-511. Markus, L. (1990) Toward a "Critical Mass" theory of interactive media. In Fulk and Steinfield (Eds.) Organizations and Communication Technology, Sage Newbury Park, CA. Rafaeli, S. (1986) The electronic bulletin board. Computers and the Social Sciences, 2, 123-136. Rafaeli, S. (1988) Interactivity: From new media to communication. Hawkins, Pingree and Weimann (Eds.) Advancing Communication Sciences: Sage Annual Review of Communication Research, Vol 16, Beverly Hills CA: Sage, pp 110-134. Rafaeli, S. (1990) Interacting with media: Parasocial interaction and real interaction. B.D. Ruben and L.A. Lievrouw (Eds.) Information and Behavior, Vol. 3, 125-181. New Brunswick NJ: Transaction Books. Rafaeli, S. and Ritchie, L.D. (1991) On the Value of Information. Paper presented at the annual conference of the International Communication Association, Chicago, May. Rogers, E.M. (1990) New Communication Technologies and the marketplace of ideas. Paper presented at the Association for Education in Journalism and Mass Communication, Minneapolis, August. Rogers, E.M. and Rafaeli, S. (1985) Computers and Communication. In B.D. Ruben (Ed.) Information and Behavior, Vol 1, New Brunswick, NJ: Transaction Books, pp 135-155. Steinfield, C.W. and Fulk, J. (1988) Computer Mediated Communication Systems as Mass Communication Media. Paper presented to the Telecommunication Policy Research Conference, Airlie House, VA, October. Swift, C.R. (1989) Audience activity in computer mediated communication. Unpublished Ph.D. dissertation, Indiana University. Thorn, B.K. and Connolly, T. (1987) Discretionary data bases: A theory and some experimental findings. Communication Research 14(5), 512-528. Thorn, B.K. and Connolly, T. (1990) Discretionary data bases. In C.W. Steinfield and J. Fulk (Eds.) Perspectives on Organizations and New Information Technology, Beverly Hills: Sage. USA Today (1990) "Computer modems as a mainstream meeting place", October 30, p. 4D.